1

An ESRC Research Group

Knowledge generation for poverty reduction within donor organizations

GPRG-WPS-023

Anthony Bebbington and Armando Barrientos

Global Poverty Research Group

Website: http://www.gprg.org/

The support of the Economic and Social Research Council (ESRC) is gratefully acknowledged. The work was part of the

programme of the ESRC Global Poverty Research Group.

2

Knowledge generation for poverty reduction within donor organizations

Anthony Bebbington Armando Barrientos

Global Poverty Research Group and

Institute for Development Policy and Management University of Manchester, UK

Abstract

The Millennium Development Goals have placed poverty reduction at the top of donor organisations� list of priorities. Failure to achieve the MDGs will no doubt place the spotlight on their financial commitments to poverty reduction, but effective poverty reduction relies as much on knowledge of poverty and poverty reduction and the kinds of knowledge that emerge from, and inform, the activities of donor organisations. The paper considers this issue in the context of Indonesia. It examines the main parameters of knowledge generation for poverty reduction in the country office of a donor organisation. It finds that organisational, environmental, intellectual and experiential factors have significance in this context.

6 July 2005 Address for correspondence: Prof. Anthony Bebbington, Institute for Development Policy and Management, The University of Manchester, Harold Hankins Building, Oxford Road, Manchester, M13 9QH, United Kingdom. Phone: +44 (0)161 275 0422 Fax: +44 (0)161 273 8829 E-mail: anthony.bebbington@manchester.ac.uk © Anthony Bebbington and Armando Barrientos July 2005

3

INTRODUCTION The Millennium Development Goals have moved poverty reduction to the top of donor organisations� list of priorities. Failure to achieve the MDGs will no doubt place the spotlight on their financial commitment to the goals, but effective poverty reduction relies as much on the knowledge of poverty and poverty reduction that emerges from, and informs, the activities of donor organisations in developing countries as it does on financial resources. Such knowledge is central to defining the parameters and ambition of the poverty reduction strategies of donor organisations. How then do donor organisations generate knowledge about poverty and poverty reduction in the countries in which they work? How do they learn from local networks sustaining both research and action on poverty? How do they use this knowledge in their operations? Notwithstanding the slowly growing literature on the ethnography of aid organizations, these issues have rarely been tackled directly, although we can presume there is a scattered literature on these issues in internal reports of donor organisations. The paper considers these questions in the context of the World Bank�s Country Programme in Indonesia. The aim of the paper is not to produce a generalisable set of findings, but instead to provide an insight into how these questions can be answered within the context of a specific case study. A case study approach is particularly legitimate for asking such questions, because the kind of knowledge generated by donor organisations operating in developing countries addresses local issues and conditions, is to an important extent, locally produced, and the parameters for knowledge generation are influenced strongly by local organisational features. A case study approach is more likely to identify the main features of such country level knowledge generation on poverty. Another important reason for adopting this approach is to throw light on issues seldom illuminated in the general literature on donor organisations. Whilst there is a great deal of literature touching on the perspective and approach of particular donor organisations at the macro- level, there is much less information on the extent to which macro and micro level factors combine to determine the way in which donor organisations operating in-country produce local knowledge. The process of knowledge generation in �satellite� parts of donor organisations may, and perhaps should, diverge in important ways from the same process in the �metro� parts of donor organisations. At the very least, this is an important issue to explore. This paper draws on work, conducted during 2003 and 2004, that sought to understand how the country programme of one donor organization (the World Bank) that is deeply committed to the MDGs generated knowledge on poverty and poverty reduction in that country, and the ways in which that knowledge production interacted with project operations.1 The work was conducted in Indonesia, one of the World Bank�s most important country programmes and a country with which the institution has a longstanding relationship � a relationship that has enjoyed periods both of great success and of serious reversal in poverty reduction. It is also an atypical country programme in that analytical work on poverty reduction has not been dominated by (largely economist) staff working within the Bank�s Poverty Reduction and Economic 1 We are grateful to the support of many people as we did this work, while acknowledging that not all agreed with our conclusions. In particular our thanks to Jehan Arulpragasam, Palmira Bachtiar, Leni Dharmawan, Richard Edwards, Donata Garrasi, Scott Guggenheim, Peter Owen and Andrew Steer. A longer version of the argument can be found in Bebbington and Barrientos (2004).

4

Management network but rather by a broader group of social scientists and economists working in different parts of the institution. It therefore lends itself to a reflection on the implications of different organizational, intellectual and social entry points for generating knowledge on poverty in a given country. The paper is divided into three sections. The first section provides background information on poverty trends in Indonesia and traces the evolution of the approach to poverty in our selected donor organisation. The second section analyses the main sources and determinants of knowledge generation on poverty in our selected donor organisation. The last section focuses on identifying the linkages existing between knowledge and operations. A set of conclusions complete the paper. 1. POVERTY AND POVERTY REDUCTION IN INDONESIA Recent trends in poverty in Indonesia In the period 1976 to 1996 there was a significant reduction in the proportion of the Indonesian population living below the poverty line, from 40.1% in 1976 to 11.3% in 1996 (World Bank for the Consultative Group on Indonesia, 2003). It is generally agreed that the main factor behind this impressive downward trend in poverty was the sustained and strong growth achieved by the Indonesian economy during this period. The financial crisis in 1997 threatened to undo the gains in poverty reduction of the previous two decades. A consistent set of poverty rate estimates for the period February 1996 to February 2002 shows the poverty headcount rate nearly doubling in the aftermath of the crisis, from 15.3% in February 1997 to 33.2% in November 1998 (Suryahadi et al., 2003), followed by a gradual decline in the recovery phase to 10.1% in February 2001, only to rise again to 13.1% in February 2002. The rapid rise in the poverty headcount rate in 1998 is explained by the 13.7% contraction of GDP in that year. The near doubling in the poverty headcount associated with the economic crisis led to a proactive examination of poverty reduction. It was no longer tenable to expect that setting the conditions for continuous economic growth was sufficient to ensure sustained poverty reduction. It also drew attention to the vulnerability of large sections of the population to abrupt changes in economic conditions. Using US$ 2 a day as a threshold indicating vulnerability, WB-I estimates that 53.4% percent of the population find themselves in this category, having declined from 65.1% in 1999 (World Bank for CGI, 2003). The high incidence of vulnerability in Indonesia has been confirmed by studies examining household survey data (Chaudhuri et al., 2001) and helped underline the direct linkages between poverty reduction and social protection.2 The rapidly unfolding decentralisation process initiated in 2001 has focused attention on the significant disparities in the incidence and extent of poverty across provinces. The headcount poverty rate has been estimated at less than 4% in DKI Jakarta, but around 55% in Papua (Government of Indonesia, nd). The national poverty map for Indonesia will also provide information on disparities at district and sub-district levels.

2 Headcount poverty indicators say little about how poor the poor are. This is an area which has not received sufficient attention in Indonesia, and is important in identifying the persistent poor.

5

Poverty is multi-dimensioned, and the MDGs will continue to bring to the fore non-income indicators of poverty and deprivation. Aggregate outcome data on the MDGs shows significant deficits in Indonesia. World Bank Indonesia noted that �some 53 percent of Indonesians lack access to at least one of the basic services� (World Bank for the CGI, 2003). Indonesia shows some progress towards achieving some of the MDGs, but progress lags in others (World Bank, 2003). Again, the decentralisation process will draw attention to disparities in non-income poverty indicators across and within provinces. The evolution of the World Bank Indonesia Office�s approach to poverty reduction The Bank�s approach to poverty in Indonesia has changed steadily over the past decade. From the mid-1960s to the mid-1990s �World Bank involvement in the country�s development efforts was pervasive� (Picciotto, 1999) with a particular emphasis on maximising economic growth. The Country Assistance Note of 1999 notes that prior strategy had focussed in particular on: boosting agricultural production, and improvements to power, transport, urban and village infrastructure, health services and skills development � along with significant adjustment lending (OED, 1999). While it was assumed that such interventions would reduce poverty, the approach was a trickle down one: from growth to poverty reduction. Indeed, the approach was in this sense remarkably successful, and per capita GNP passed $1000 p.a. in the mid-1990s. However, the institutional (as opposed to infrastructural) foundations of this growth received less attention. While the Bank talked euphemistically of a �high cost economy,� �leakage,� and �inadequate accounting� (World Bank, 1999), corruption and theft were not addressed openly or directly (Friend, 2003). Remarkable growth, leading to remarkable poverty reduction, led to a certain silence on deeply problematic processes of governance � and it is probably the case that even if the Bank was aware of them, it neither understood them nor their form of operation. The approach up until the crisis was a conservative version of the WDR 1990 approach to poverty reduction: growth plus human development and (in this case limited) social protection. One consequence was that poverty dynamics received relatively little attention, and the non-income dimensions of poverty even less. The analytical agenda focused more on macro-economic and public expenditure issues. Work within the sectors proceeded apace but largely to support Government of Indonesia demands for top-down managed service and infrastructure delivery. Sectors were not primarily concerned with the links between projects and poverty dynamics, and generally lacked specific analysis of these dynamics. The crisis of 1997 made it evident that this was not enough. A rather despairing CAS Progress Report (World Bank, 1999) identifies �institutional policy� as the �great failure� of the New Order Government � and relative silence on problems of corruption, governance and financial sector regulation problems as the Bank�s.3 Likewise the strategy right up to the CAS of 1997 was criticized for its inattention to poverty (OED, 1999). The crisis thus pushed governance (primarily in the form of

3 �Where we failed was in not pushing harder the message that sustainable development requires not only good economic policies, but good institutions as well� (World Bank, 1999: 8).

6

corruption) and poverty much more forcefully onto the agenda. By 1999 the emphasis on infrastructure was temporarily abandoned,4 macroeconomic stability meant work on financial and corporate restructuring, and the stress was on protecting the poor, improving governance and combatting corruption. This manifested itself in a dramatic shift in the country portfolio as the Bank scrambled to find instruments to deal with poverty and crisis. Two very large new loans, the Social Safety Net Adjustment Loan (SSNAL), and the Kecamatan Development Project (KDP), were approved as means (albeit quite different in approach) of targeting the poor while also injecting significant resources into the public budget in order, inter alia, to keep the economy afloat and so create conditions under which democratic elections might go ahead. In some sense the completely distinct experiences of these two loans have determined the subsequent country programme and its approach to poverty. As acknowledged by the bank, SSNAL was a deeply problematic operation: �We cannot emphasize enough the difficulties we are facing in designing the loan in a way that it ensures good governance in public expenditure management while at the same time meeting the immediate resource needs of the economy� (World Bank, 1999: 10). These difficulties were never overcome and in the face of massive robbery and difficult disbursement because of highly specified (and some in the Bank would say cumbersome and futile)5 targeting rules, the loan was cancelled. Meanwhile KDP became, in the eyes of some in the Bank, the only part of the Bank�s portfolio that worked. While there was also corruption here (as KDP itself documented: Woodhouse, 2001), it was more decentralized and far less pervasive. Meanwhile villages lauded the project, it fostered participatory planning and resource allocation on a massive scale, and (importantly given both the Bank�s internal incentives, as well as the need to prop up the Indonesian economy) it disbursed large amounts of money ahead of schedule. KDP became a touchstone across the Bank for how to do community driven development on the sort of very large scale the institution values, and it has subsequently appeared in World Development Reports (e.g. World Bank, 2001a; 2004) as a model of how to link participatory governance and poverty reduction, and how to foster empowerment. Whether it merited quite such unqualified accolades is another question. The trauma of SSNAL plus the crisis left some in the country office convinced that the Bank should lend little to the Government of Indonesia; and left most convinced that whatever the case governance issues (largely taken to mean corruption) had to be central to a country strategy. Meanwhile the apparent success of KDP suggested that intervening through decentralized levels of government, and coupling that intervention with large scale, externally induced forms of participation, was perhaps a viable strategy on which the Bank could focus its energies. Under the influence both of these country office experiences as well as of the debates around the WDR 2000/2001, governance and poverty reduction became coupled.

4 The CAS progress report refers to a temporary abandonment, but it is now 5 years in (World Bank, 1999: 9). 5 �Futile� because of the inherent fungibility of such loan resources. These concerns about excessively specific and unhelpful targeting conditions for disbursement may play themselves out again in current discussions about how specific Indonesia Local Governance Reform Project�s conditions should be. Some in the Bank appear to favour more specific conditions to guard against corruption; others favour less specific conditions for reasons of fungibility and to guard against a failure to disburse.

7

This is captured in the claim of the 2001 report on poverty reduction in Indonesia that �a poverty agenda is a governance agenda� (World Bank, 2001b). The 2002 update on the CAS for 2001-2003 likewise emphasizes governance: �Indonesia�s success or failure in building open and fair governance institutions � both national and regional � continues to be the most important determinant of the longer term prospects for Indonesians.� (World Bank, 2002: 4). Strategically these lessons and reflections translated into a country programme that in the early 2000s dramatically reduced its lending to Indonesia6 on the grounds that governance problems made greater lending inappropriate. With this reduced lending, the programme focused attention on: fostering macroeconomic stability and growth; promoting accountable government; and delivering better public services to the poor (World Bank, 2002: 10). The shift from the CAS of just 4 years before was dramatic: promoting good governance and improved services for the poor become two of the three main pillars of the Bank�s approach. Operationally, the lending programme was dominated by KDP (which entered a second phase), an emerging urban version of KDP (called UPP) and other ongoing projects. Adjustment lending (including the SSNAL) was suspended. The 2004 CAS constitutes an elaboration of the CAS 2001-2003 refracted through continuing change in the national context and influences from wider discussions within the Bank. The national context was one of: continuing, but sluggish growth; of deepening decentralization to district and urban government; and of growing disenchantment and disappointment with the reform process in Indonesia in the face of continuing corruption. The intellectual influence of the wider Bank comes from the debates around and thinking behind two WDRs: that of 2000/1 on Attacking Poverty, and of 2003/4 on Making Services Work for the Poor. WDR 2004�s influence is evident in the CAS�s heavy emphasis on enhanced public involvement in driving human development services. The intellectual influence of debates around WDR 2000/2001 is less direct but is apparent, for instance, in how the demand (empowerment and voice) and supply (public institutions, justice etc.) sides of governance are discussed. This said, it is also the case that this influence has worked in both directions for some members of the Indonesia country team have at certain points influenced thinking in these WDRs, both as WDR team members, and as participants in the networks of discussion surrounding, the writing of these reports. In this sense, the WDRs reflect Indonesian experiences as much as the other way round. While both WDR 2000/1 and the CAS talk of working on both the supply side (making institutions more accountable, transparent and accessible) and the demand side (increasing demand for such qualities), there is a difference of emphasis in how demand is discussed in the two documents. The CAS talks of greater public participation in policy making and monitoring; WDR 2000/2001 (World Bank, 2001a) talks also of supporting organizations of the poor, social movements etc. The advantage of engaging with such movements and organizations is of course that this would broaden the Bank�s understanding of poverty and governance in Indonesia. Also, to the extent that the Government of Indonesia will prioritize poverty reduction only when there is significant societal demand that it do so, building the capacity of

6 By 2002 the Bank was lending at less than a third of its historical levels of lending, and had ceased adjustment lending �until conditions are more propitious� (World Bank, 2002: 5).

8

such organizations to make analytically well-grounded demands is an important task in itself. The new CAS constitutes a compromise between the push to lend and the caution based on the experience with the Indonesian crisis and the scale of corruption in the country. By foregrounding governance, and insisting that the issue is less how much is lent, but how lending decisions are made � namely only after good governance criteria have been satisfied � the CAS achieved a compromise which made lending conditional on evidence that the quality of governance had improved. This of course also has implications for any programme of knowledge generation on poverty in Indonesia, because the emphasis on improving governance as a means of fostering poverty reduction makes understanding the ways in which governance and poverty dynamics are coupled critical (as foreshadowed in the 2001 poverty report). The CAS rests on the argument that improving governance is central to achieving its two main objectives of improving the investment climate (which would, inter alia, create more income opportunities for the poor) and making service delivery responsive to the poor.7 It also takes seriously the aid effectiveness argument (i.e. that aid is only effective where governance is good), and aims to operationalize it at different scales. It thus seeks this improved governance at four levels around which the lending programme is to be arranged: the community/sub-district, the district, the national, and among public service utilities (especially water). Of these the district (kabupaten) level is deemed the most critical for several reasons, and this focus on the district marks an important change from previous CASs. The kabupaten focus is important not only because of its prominence in decentralization legislation but also because to focus lending only on those kabupatens demonstrating reformism introduces a degree of competition in which districts might be encouraged to foster good governance as a means of accessing Bank support (an incentive to good government that a focus on central government did not allow). The CAS argues that good governance in Indonesia will translate into poverty reduction through enhanced human development (via improved services) and increased income (through growth triggered by new investment). In this regard it is more strategically poverty oriented than the 2001-2003 CAS which � though it emphasized poverty � was primarily concerned with dealing with the poverty effects of the crisis. The new CAS takes a more medium term view. As such it creates spaces for poverty reduction work in ways that previous CAS�s have not. In placing the stress on governance, the CAS points to an important area for analytical work on poverty and poverty reduction: the specific causal mechanisms linking dimensions of enhanced governance to particular types of growth and service delivery effects, and their poverty reduction properties. While there is some work elsewhere that may suggest the nature of these causal pathways, there is very little work on the nature of such pathways in the Indonesian context, especially at a district level. While the Bank has been at the forefront of work exploring governance-poverty linkages at a village/neighbourhood level (see for instance work conducted around KDP and

7 With this dual focus on investment climate and human development services one might argue that the new CAS is something of a return to the mid 1990s� emphases on growth and human development. The way in which these objectives are sought is, however, radically different � no longer through central government and no longer without scrutiny of governance processes.

9

UPP),8 it has done much less at kabupaten and higher levels. Here there is scope for combining case studies aimed at unpicking causal pathways and larger comparative studies aimed at telling larger (if less nuanced) stories. Such studies could usefully focus on how specific governance improvements translate into more pro-poor forms of growth, and more pro-poor service delivery systems. Understandings of poverty within �satellite� donor organisations It is commonplace that, within country offices in donor organisations, differing views of poverty, its primary causes, and effective remedies, coexist � and in the Bank such differences were also apparent. To an important extent, these views on poverty and its reduction policy are not mutually exclusive, and are sometimes indicative of different emphases of interpretation that manifest themselves in operational decisions and in documentation. We might characterize these views, in no particular order, as:

o poverty is primarily a problem of assets and poverty reduction strategies ought to aim to provide those assets more effectively

o poverty is primarily an effect of service delivery, especially of key services such as heath, education, social protection

o poverty reduction is primarily a governance problem and interventions should focus on enhancing the quality of governance

o poverty reduction is primarily a problem of social exclusion and interventions ought to expand the scope for poor people to participate in the institutions, practices and benefits of society

o poverty reduction is primarily a problem of distribution and of social power, and interventions should aim to redistribute both power and assets

o poverty has once again declined significantly in Indonesia and can be addressed through broad development policies; and relatedly

o poverty is a problem of growth Most country office staff are likely to identify with all these views on poverty � but they would emphasise one more than the other, and would also be aware of the sensitivity associated with certain viewpoints, within and outside the Bank. Such a breadth of emphases is healthy and can contribute to debate and learning on poverty reduction. The view that poverty is still primarily an asset problem is one that stresses the continuing importance of basic services and infrastructure to poverty reduction. It does not deny that governance is a real problem in Indonesia, but it does want to insist that the success of projects must still be assessed in terms of their impact on the asset bases of poor people. Therefore, it continues to insist that the Bank not lose sight of the critical importance of improving the quality of health and education services, of translating those services into human development impacts among the poor, and of certain forms of infrastructure provision. Theories of basic needs and entitlements to human development are very relevant to this viewpoint. Likewise the view that governance is the central problem in poverty reduction does not naysay the continuing importance of quality service delivery. However, it is a

8 See for example Grootaert (1999) and Chandrakirana (1998).

10

view more inclined to argue that in the absence of improved governance arrangements any attempt to improve services in Indonesia will be limited in its impact and sustainability. This is a view that draws its intellectual inspiration from institutionalist approaches in growth theory that insist that the factor in economic development is the quality of institutions, as well as from certain approaches to state-civil society relationships. The social exclusion view goes deeper still in its explanation of poverty. This view would seek to explain both problems of governance and inequities in access to services in terms of underlying social relationships that systematically exclude significant segments of Indonesia�s citizenry. This view draws inspiration from those intellectual (and political) traditions that emphasise social structures in their explanations of development; normatively it also draws on human rights traditions. Many expressing this latter view would go a further step and suggest that poverty in Indonesia will only be addressed if assets (including land) are redistributed, and if the power of the military and other elites is reduced. They would say that the problem is not so much one of income poverty, but rather of power relationships and social inequality.9 These views are less prominent in the Bank�s policy and programmatic statements, but more prominent among important bodies of academic writing on Indonesia, especially from non-economists. These views, not surprisingly perhaps, also reflect the range of views widely discussed in the development literature, as well as an understanding, in many cases drawing on personal experience and reflection, of the types of intervention strategy that are likely to be effective in Indonesia. It can be argued that such a coexistence of different perspectives on poverty does not generate incoherence at the operational level because governance, rather than poverty reduction, has been the dominant concern in the country office operations. Whatever the case the stronger focus on poverty reduction that appears to be emerging within the programme (and that is being promoted by some of the Bank's bilateral partners, such as DfID) has inevitably generated discussion about the most effective strategies for poverty reduction in Indonesia. It is also helping drive a more extensive monitoring and evaluation of programmes, and the need will arise to define and articulate more precisely poverty perspectives within the country office.10 2. GENERATING FORMAL KNOWLEDGE ON POVERTY REDUCTION An increasing concern to fashion programmes addressed at poverty reduction necessarily requires bodies of knowledge on which such programmes will draw, Such knowledge, however, can derive from a range of sources � formal and informal, in-house and external, national and international. This section explores some of the

9 A view that will receive more intellectual and political support from the forthcoming WDR on Equity and Development (World Bank, 2005). 10 In the context of current Indonesian conditions, it would make a great deal of sense for a �poverty oriented� country programme to focus attention on vulnerability, the likelihood that people will find themselves in poverty. It is also important to focus analytical and policy attention on those suffering from persistent poverty, who are the most vulnerable. Likewise knowledge deficits associated with how the dynamics of vulnerability (and wealth formation) affect different types of poverty, would become more apparent.

11

ways in which the Bank's country programme in Indonesia has attempted to generate and access such knowledge, and the (intellectual, burueaucratic and developmental) reasoning underlying these differing strategies. In-house generation of knowledge Donor organisations are in a position to generate formal knowledge on poverty through both their research and operations.11 The bulk of the World Bank Indonesia Office�s analytical work on poverty between the mid-1990s and early 2000s had been conducted within its social development sector. This is quite unlike the situation in most other Bank country offices, and occurred for quite particular reasons. These created an environment in which (largely because of the character and resource mobilizing skills of the lead social scientist, and later the operational success of KDP) the Social Development group occupied a central role in generating knowledge on poverty and � from that position of relative strength � brought non-economists and economists into the same team, working on a shared set of analytical problems themselves often closely linked to operations. In this process, Social Development became home to a team of non-economists and economists that, while they had operational responsibilities also (a) had a clear analytical bent, (b) were committed to combining qualitative and quantitative work, and (c) became progressively more convinced that a community based approach ought to lie at the core of an effective poverty agenda. Indeed, the current state of knowledge of poverty in the country office has been shaped, to a large extent by two Social Development studies. The first was the Local Level Institutions study (LLI), conducted between 1996 and 1998 (LLI1) and then repeated between 2000 and 2002 (LLI2). Local Level Institutions Study-1 and the insights it offered regarding the relationships between village level collective action and household welfare underlay the design of KDP and it is likely the case that without LLI1, KDP would not have happened (Guggenheim, 2003; but see below).12 Each Local Level Institutions Study was heavily field-based (conducted in 48 villages in three provinces) and combined qualitative and quantitative analysis. The Local Level Institutions Study was also significant because it constituted one of a very select number of significant field studies exploring the links between poverty and social organization in Indonesia that were conducted during the 1980s and 1990s. The second study was the 2001 report on poverty in Indonesia (World Bank, 2001b). This drew heavily on Local Level Institutions Study as well as on two field based initiatives designed to generate real time information on poverty and social dynamics in Indonesia. In each case, the rationale for this heavily field oriented approach to developing economic data was that there was no other system in place to monitor what was happening in provincial Indonesia, and that government statistics were not 11 By �formal� we mean analysed, organized, documented and (often) published in one form or another (if only in the form of stapled overheads that then circulated widely and took on a life of their own: Chandrakirana, 1998). Bank staff also learn from operations and the process of doing research � we treat that type of knowledge differently (in a following section) on the grounds that it is less visible, less organized and less processed, though certainly often more important to the bearers of the knowledge. 12 For reasons of transparency it should be noted that Bebbington was involved in LLI at its very beginning, and then again in 2002 in an effort to conceptualize some of its findings for publication in peer-reviewed journals (Bebbington, Dharmawan et al., 2004; forthcoming).

12

sufficiently robust. At least one of the studies, coordinated by a Bank economist, was based on the methodological commitment that in monitoring poverty and service delivery in Indonesia �ninety percent of the battle is just sending an honest and sincere person who finds out what was going on.�13 The first of the two studies monitored the inter-provincial trade conditionalities of an adjustment loan in the late 1990s (some of this data was later reported in Sumarto et al., 2003). The second body of work monitored the ways in which people were coping with the crisis of the late 1990s and how well government programmes were reaching the poor. Each initiative involved placing sizeable teams of researchers in the field to monitor local processes. These researchers reported back periodically and their data was then analyzed more formally. Both the Local Level Institutions Study and the poverty report consciously aimed to shift the poverty agenda in the Bank�s country programme by arguing that poverty reduction was more effectively advanced through a governance agenda and by insisting on the multidimensionality of both local livelihoods and the experience of poverty. Each involved large teams of field based researchers well networked into village level social processes, with a relatively small centralized team of researchers with expertise in particular analytical techniques. This contrasts with the more recent organization of poverty analytical work where the emphasis on experts is greater and that on embedded field researchers less. Subsequent analytical work14 in Social Development has been much more closely linked to operations � specifically KDP. This model of knowledge generation attempts to understand and generalise from specific programmes. It is an �organic� process of knowledge generation, broadly participative, and embedded in the researchers. The clear weakness of such an approach has been that the write up and dissemination of the knowledge generated is compromised by operational demands on staff time, and constraints on research time. As a result, while people in the country office are aware of studies such as Local Level Institutions Study and their broad conclusions, they know far less about the details of their findings. Generation of knowledge about poverty within PREM has differed in several respects. Typically PREM�s analytical work in Indonesia had dealt primarily with macro-economics, public sector management and country wide statistical analysis and drew little on operational experience. More recently � and reflecting a heavy injection of bilateral resources specifically addressed to support the poverty orientation of the Bank's country programme � PREM researchers have paid increasing attention to poverty questions, and some of this work has covered the analysis of micro-data, for example in the analysis of the incidence of poverty and deprivation, the construction of poverty profiles, and in poverty mappings. This model of knowledge generation involves outside partnerships with government agencies in the design of surveys and collection of data, as well as partnerships with other parts of the Bank and independent researchers in the poverty analysis. This research has been extractive and quantitative, and the output feeds directly into country office set-pieces and more broadly into the Bank�s poverty work. 13 We can speculate that such an approach could have been more easily agreed in Social Development than in PREM where methodological protocols would arguably have been more demanding. 14 This includes work on conflict; research on poor people�s access to justice; and other smaller pieces of work.

13

These different approaches to knowledge generation about poverty and poverty reduction point to real potential synergies. Neither approach is sufficient by itself. Each can complement the other in important ways. Partnerships for knowledge generation Just as it was protest from Indonesian society (rather than any pressure from the Bank) that finally led a new Indonesian government to begin talking about corruption, so it is likely that the government will only place poverty reduction at the centre of its policy agenda if and when society demands this. In this sense, any effort to strengthen Indonesian research capacity to generate knowledge on poverty in such a way that enhances public debate should be viewed as a development intervention as much as an academic one. Likewise strengthening broader civil society ability and interest to engage in public debate on poverty is equally important. These challenges are all the more important given the small size and limited capacity of the Indonesian university and civil society research community � a small size that is itself an effect of New Order styles of authoritarianism. At issue here, then, is the extent to which country offices of donor organisations successfully generate knowledge jointly with other actors in such a way that builds research capacity. Many factors influence the likelihood that this will happen, though here we focus on two. First, the quality of the relationships existing between �satellite� and �metro� levels of a donor organisation are significant in facilitating, or obstructing, local capacity building � in particular through determining the incentive structures for Bank staff. Second, and related to this, the time horizons with which Bank staff operate (in particular those in the country office) will also have great influence on the potential for capacity building. Where staff operative with a short-term horizon, efforts directed at local research capacity building may appear less productive if these yield results only in the longer term. Similarly, professional incentives within donor organisations may ensure that a capacity building approach to knowledge generation is insufficiently rewarding and deemed unlikely to generate work of a quality that will carry weight within the organisation.15 In certain donor environments, incentives and structures have been such as to encourage such capacity building. Hofman et al. (2004) have noted that such capacity has been created in the past in Indonesia. They refer to the way in which the Ford Foundation helped create cadres of economists who went on to play critical roles in ensuring effective economic management within the Indonesian government during important periods of the New Order. This process demanded sustained donor commitment over many years, and indeed the quality and professional capacity issues noted above go hand in hand with questions of timing. If it will take time to build quality, results may not be available in time for planned Bank operations.

15 One doesn�t need to go as far as S.Landsburg�s �people respond to incentives, the rest is commentary� (cited in Easterly, 2001, p.xii) to agree that incentives are relevant here. Country office staff in donor organisations will be much more likely to dedicate themselves to eliciting such local capacity the greater the professional incentives to do so. Yet the pattern of incentives may be stacked up against longer term, cross-network activity such as developing local capacity. The careers of Bank staff depend as much on the incentive structures within their respective Bank-wide networks (PREM, Social Development, HD etc.) as on those within the country programme.

14

That said, research capacity building has been possible even within existing incentive structures. A significant example of recent successful poverty research capacity building in Indonesia has been the emergence of the independent research institute SMERU, born to a considerable extent out of earlier poverty work conducted by Social Development in the context of the Indonesian crisis. SMERU is one of very few Indonesian research institutes dedicated to analyzing poverty and development, and plays an important role in challenging government statistics, in disseminating information on poverty to Indonesian society, in fostering poverty mapping and other work. However, SMERU did not emerge so much because Bank incentive systems directly encouraged it. Rather, it emerged as a by-product of other poverty analysis and project monitoring work in the Bank and as a result of specific initiatives of certain Indonesian and Bank researchers. Either way, the growth of SMERU does demonstrate that under certain conditions an environment can be created that helps high quality independent research institutes to emerge and become consolidated.16 Learning about poverty reduction Donor organisations learn from a range of sources. These include: the accumulated experience of staff; research conducted elsewhere in the donor organisation; operations; and research and insights of other donors, academics, independent research institutions and civil society in Indonesia; and the research and operations of government departments. The relative weight of each source varies among the different parts of a country office. Some of these sources are solely experiential, others are solely analytical, and most are a mixture. In the World Bank Indonesia Country Office, and we suspect World Bank Country Offices more generally, great weight is placed on the experience (research and operational) of staff as a source of insight into the dynamics of poverty and how to address this poverty. There are many reasons why such practical knowledge is valued so highly. First and foremost is its inherent value. Second, a country office is as much a team of practitioners as it is of researchers, and this influences the types of knowledge that have most credibility within the office. Third, people�s time to read analytical work is greatly constrained. Staff also learn from operations. Sometimes this is formalised within the project design, where these set aside funds that can be used for specific studies and more analytical forms of monitoring. In the Indonesia country programme, the extent to which operations have explicitly incorporated formal learning processes on poverty varies across sectors and over time. In the mid-late 90s, operations did little in this regard. Even if poverty issues came explicitly into projects, in the view of one senior staffer it was not uncommon that �they didn�t have the data on the points they were trying to make.� �They were working with the ministries and how the ministries thought,� hence the absence of evidence or reflection on poverty dynamics was not surprising. The Bank�s approach was consonant with that of the Government of Indonesia. It is not clear how far this has changed in the early 2000s. Operational staff�s limited available time has been taken up in rethinking projects in the post-crisis 16 The centre Akatika in Bandung is another case of a centre that emerged from an academic research project (in that case led by the ISS in Holland) and that is now able to do quality, provincial level analysis (Antlov, pers. com.).

15

period which has meant that commissioning analytical work on poverty in their sector has not been a priority (though there are clear exceptions). Another reason why operations have done little analytical work on poverty in their sector may be that they have had to devote so much more attention to dealing with corruption issues (as well as simply getting projects implemented and supervised). A further reason is that at a sector level, there is still a clear concern for sector specific issues that, in the face of the need to resolve operational problems and respond to policy change, often gain more attention than thematic concerns such as poverty and livelihood dynamics (World Bank, 2004a). While the tendency is still to rely on broad assumptions and stylised facts regarding the linkages between poverty on the one hand and vulnerability, growth, and governance in the other, there is widespread recognition that analytical work is needed to clarify these and maximise programme effectiveness, and that operations provide a fertile ground in which to do this work. More work needs to be done, most of it of an analytical nature, to pin down and measure these broad linkages. There is also some academic research on poverty, poverty reduction, livelihoods and governance in Indonesia. However, this literature is not invoked by Bank documents as often as it might be,17 leading to concerns that some analytical work in the country office might be simply reproducing findings that are already known. The significance of civil society for the World Bank Indonesia Office�s understandings of poverty has waxed and waned over the years. There are both formal and informal mechanisms through which donor organisations can learn from civil society. The formal mechanisms include organised consultations with civil society organisations and meetings with invited experts. Informal mechanisms derive largely from the social networks within which staff members participate. These networks provide a measure of how deeply (and where) country office staff are embedded in society, and thus also of how staff are tapped into the experiential evidence that that society produces about poverty and development. In Indonesia, some Bank staff have quite strong links with activist professionals and NGOs (although far weaker links with social movements, labour organizations and the like). The numbers involved, however, are still small. Yet, the extent to which staff of donor organisations are embedded in the society in which they work is key to facilitating these linkages.18 The extent to which donor organisations are able to reach outside the metropolitan centres and learn about local socio-economic processes is also relevant in this regard, given the extent to which Bank supported programmes reach across all parts of Indonesia.19 Learning about these processes presupposes a level of embeddedness in Indonesia�s local institutions. Of course, this is not easy and it is not surprising that Jakarta-based staff might find it difficult to learn about the microdynamics of change in a community in South Sulawesi, or in a kabupaten capital in Kalimantan; and it is

17 For a notable exception, see Hofman et al. (2004). 18 Staff who are conversant in Bahasa Indonesia are far more able to develop links within Indonesian society; those who are not are clearly disadvantaged. The aggregate effect may be that expatriate donor staff talk primarily to other expatriate donor staff. This is an issue for many donors. As noted by the country representative of one agency,: �one gets the sense that donors are spinning their wheels off there by themselves � while everything else just carries on.� 19 For all its simplicity this is a theme that has emerged in other of our research on World Bank programs that aim to foster empowerment (Bebbington, Lewis et al., 2004).

16

perhaps not even surprising that learning about the microdynamics of policy formation in Jakarta itself is difficult for such a team. But these microdynamics clearly matter to the success of projects and of policies. For instance, whether intra-community relationships and dynamics at the interface of health service provider and user mean that health care vouchers are serving primarily to help poor people access health services, or primarily to help service providers supplement their salary, matters to the poverty impact and orientation of health services. Whether business elites in a kabupaten capital are able to leverage a share of education reform resources so that they are channelled to private schools owned by and used for the families of these elites also matters. It is difficult for short project supervision visits to pick up on such issues, not least because the combined effects of linguistic diversity (with at times three or four languages at play) and local efforts to stage manage monitoring visits tax even the most skilled monitor. It would also be impossible and unwieldy to trace such dynamics in all sites where projects operate. However, understanding at least some of these dynamics can offer very useful information for designing incentive and accountability mechanisms in projects. To generate such information would be a slow process, depending largely on ethnographic methods. But it need not be a costly process. Yet such ways of generating a different type of knowledge about the dynamics of change and poverty reduction in Indonesia are not yet well developed in the Bank�s approach to learning about Indonesia. The general issue in the preceding discussion is how to secure a set of incentives and a culture within donor organisations that encourage more basic levels of learning about social and poverty dynamics, and the political economy with which projects articulate. For many donor organisations, the very pressure of work dissuades such learning � for it requires time, immersion and relatively long term research. In an environment where the emphasis is on solving project problems, contracting consultants, preparing projects and keeping up with a barrage of email, work routines genuinely obstruct serious, basic and strategic research. One could argue that they even crowd out the space that people need to be able to frame the deeper, structural questions that such research would address. As in the case of capacity building, many of these pressures and incentives derive from organisational dynamics and career structures and are thus not easily resolvable within country offices. 3. POVERTY KNOWLEDGE AND POVERTY TARGETTING In the context of the World Bank Indonesia Country Office, relatively little analytical information has been produced within operations to assist them in addressing poverty, and until recently this was not a primary concern. Even today, projects have relatively little empirical knowledge at their disposal to think through themes such as: poverty targeting; disaggregating among levels and types of poverty; the conditions under which the poor can be substantively and substantially involved in holding services to account; clarifying the particular links between governance and poverty in their sector/geographic area; and so on. Poverty targeting in Bank projects is an approximate science. Relatively little is done to ensure accurate targeting of the poor. The emphasis is instead on area based targeting: that is projects go to poor desas, poor and conflictive kabupatens or kecamatans, and so on � the unit of targeting is rarely the household. However, while

17

projects do not target the poor, they are more likely to target the vulnerable. This reflects one of the ways in which poverty is talked about in the country programme. In fact, while a quarter or so of the population live on less than $1 a day, over half live on $2/day or less. The implication is that while a significant number of people are poor, the problem of vulnerability to poverty is that much greater. The problem of chronic poverty is recognized, but not treated in any detailed way. To the extent that the vulnerable population is a priority concern of the Bank, then it becomes less important to pursue any more precise targeting of the very poor. In a country where over half the population is vulnerable, a project in a poor kabupaten, kecamatan or desa is likely to reach a significant vulnerable population. Indeed, this broader notion of the types of poverty that ought to be addressed allows for more flexible geographies in funded interventions. Moving down a scale, some projects then seek more geographically specific forms of targeting. Remoteness is taken as a surrogate of poverty in some projects (e.g. Basic Education, South Sulawesi), and in some cases (as in Basic Education) the project and its partners have developed simple GIS�s to assist this targeting. In practice, however, it was not clear how rigorously the criterion was applied (given the number of roadside schools that also benefited from the project). This general pattern is affected, and complicated, by the governance filter and governance objectives the Bank applies to its operations. Thus some projects combine targeting and governance objectives. The rural CDD component of BIUP channels funds through the kepala desa adat partly on the grounds that the traditional leader will have better knowledge of who is poor, and because channelling funds this way will strengthen traditional governance institutions. Both seem to be the case, though the traditional nature of these governance institutions seems to come with a general lack of participation in village governance because participatory decision making would be to question the traditional leader�s authority. It also leads to some spreading of benefits to the middle poor in order to preserve the leaders� legitimacy. KDP also resists household targeting on the grounds that the project aims to strengthen groups and social capital � and that to target individuals would have the effect of creating jealousies among beneficiaries and non-beneficiaries and so weaken social capital.20 This again means that some benefits are enjoyed by the middle and less poor � though the conditions regarding women�s involvement in groups and approved projects offsets some of this �leakage.� Governance affects poverty targeting in other ways. The local governance platform emphasises working with reform-minded kabupatens. These are not necessarily the poorest. While the wealthiest of these kabupatens have been excluded, the governance conditionality will likely take Indonesia Local Governance Reform Project to kabupatens that are not the poorest. So the governance filter may mean less targeting of acute, persistent poverty. On the other hand it may address a key cause of vulnerability (to the extent that poor governance aggravates vulnerability and weakens crisis response mechanisms). It is argued cogently that a significant (perhaps the main) cause of vulnerability is the overall weakness of institutions from the village through to the national civil service (World Bank, 2003; Hofman et al., 2004). If weak, corrupt and easily captured institutions really are a cause of vulnerability, then

20 Both Gonzalez de la Rocha (2003) and IFPRI (Adato, 2000) have argued the same point in pointing to the social capital weakening effects of household targeting in Progresa, Mexico.

18

a portfolio that emphasizes improving governance might be said to be automatically a strategy for reducing vulnerability. This raises the issue that more analytical work is needed to be certain that there are strong links between governance and poverty and, more importantly, on the specific causal mechanisms involved (see also Sumarto et al., 2003).21 The specific causal mechanisms would need to be identified in order to know how to address vulnerability (and other dimensions of poverty) via a governance lens. It is more than possible that the causal mechanisms implied by project designs may not be the most significant, and if so that the poverty effects of these projects may not be that great. In the absence of more detailed knowledge on governance-poverty linkages, then, it is probably not sufficient to replace a poverty agenda with a governance one on the grounds that improved governance will in the end improve poverty (though there may of course be other reasons for prioritizing a governance agenda). Stronger linkages between knowledge and operations would provide the opportunity for the knowledge generated on poverty and poverty reduction from operations to feed back into the latter, and strengthen their poverty reduction effectiveness. 4. CONCLUSIONS The paper has focused on the main parameters governing knowledge generation about poverty and poverty reduction in one donor organisation in Indonesia. The background to the present situation is informed by weaknesses in the Bank�s mode of operation and understanding of Indonesia that became painfully apparent following the Indonesian crisis. These weaknesses traced their roots to the Bank�s pre-1997 relative inattention to poverty dynamics in both its analytical and operational work; its even greater inattention to governance issues and the causal links among governance, growth and poverty; and its prioritization of lending over learning. Much has been done to remedy this situation in the post-crisis period. Greater attention has been paid to setting in place effective knowledge generation processes about poverty and poverty reduction which can exploit the linkages existing between learning and operations. These could: throw light upon governance-poverty linkages which remain poorly understood in anything other than general terms; improve the extent to which projects target poverty effectively; and embed analytical work within on-going processes of debate and change in Indonesia. The key parameters applying to the generation of local knowledge about poverty and poverty reduction continue to exercise influence. Organisational factors such as the nature of the relationships existing within donor organisations between country offices and the centre, the nature of professional incentives, and those influencing the 21 The Bank�s arguments about causality are also complicated by the poverty reduction successes of the New Order regime. On the one hand there is a desire to distance the Bank from the New Order for reputational reasons: the emphasis on corruption and governance problems exemplifies this. On the other hand, the poverty reduction effects of the New Order regime are undeniable, and the Bank recognizes these - indeed the Bank is also implicated in these successes. Yet, to recognise these effects weakens arguments that bad governance aggravates poverty, because the Suharto government was clearly not an exemplar of good governance. To argue that had there not been corruption there would have been even more poverty reduction is not necessarily convincing as it may well be that some of the social arrangements that made possible the poverty reduction effects of the New Order may have been the same ones driving corruption.